How Radio Station Music Rotation Actually Works: A Software Engineering Perspective
By the KAVANA engineering team
Music rotation sounds simple: pick a track, play it, pick the next one. In practice, it is a constrained optimization problem that has to run in real time, every day, for decades.
The Core Constraints
A music rotation system must satisfy constraints simultaneously:
- Hour count: A track should not repeat within N hours (typically 2-4 hours for hit music, 6-12 for gold)
- Artist separation: Two tracks from the same artist should not play back-to-back, or even within the same segment
- Sound consistency: Consecutive tracks should be sonically compatible (energy, tempo, key)
- Day of week variation: Monday morning should sound different from Saturday night
- Daypart targeting: Drive time (06:00-09:00, 16:00-19:00) gets the highest-rotation, highest-energy tracks
What Rotation Actually Means
Radio uses named categories:
- A rotation: Newest and hottest, plays 5-8 times per day
- B rotation: Current hits, 3-4 times per day
- C rotation: Recurrents, 1-2 times per day
- Gold: Catalog, varies widely
The scheduler maintains a queue for each category and handles promotion/demotion based on week-over-week performance data.
The Database Layer
Each track has extended metadata beyond the obvious:
TrackID VARCHAR(36)
Title VARCHAR(200)
Artist VARCHAR(200)
Duration INTEGER -- milliseconds
Energy FLOAT -- 0.0-1.0, from audio analysis
BPM FLOAT
Category VARCHAR(10) -- A/B/C/Gold
LastPlayedAt TIMESTAMP
PlayCountToday INTEGER
The scheduler joins on LastPlayedAt and PlayCountToday to enforce separation rules.
Handling the Hour Boundary
The hardest operational problem is the end of the hour. The broadcast clock has hard boundaries: news at :00:00. If the scheduler places a 4-minute track starting at :57:30, that track will still be playing when news is supposed to start.
The practical solution: schedule tracks with predicted end times, flag any track that would cross a hard boundary, and substitute a shorter track from a fill pool. The fill pool contains station IDs, jingles, and short instrumentals indexed by duration range.
The AI Extension
Modern systems augment traditional rotation with real-time context: current weather, local events, time of day. A track tagged "summer/outdoor" gets a rotation boost when the temperature exceeds 30C. This requires the rotation engine to pull live context and re-weight selection probability at runtime.
The implementation uses a base score from rotation rules, multiplied by a context-weight lookup, normalized to sum to 1.0 across the eligible pool. Context weights update on a configurable interval, typically 15-30 minutes.
KAVANA has been building broadcast scheduling and rotation software for 500+ FM stations since 2005. Learn more at kavanafm.com.
